Hi all.
Im noob here and i was trying to understand how to use and the propose of this methods.
I will appreciate examples of how make it work. Thanks all 
create_head is used to create a bunch of linear layers with some non-linearity which are used for final classification, taking the input given by the model created by create_body. create_body takes the pretrained network, cuts the linear layers from the end which are available in the pretrained network and cut the rest of the conv layers into 2 groups for discriminative learning. In the end we have input->create_body->create_head->output.
2 Likes